> ## Documentation Index
> Fetch the complete documentation index at: https://docs.wittify.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Volume Forecasting
> Plan ahead with confidence.
**Volume Forecast** is the page that projects the next 7 to 180 days of call volume. There are two places to see the forecast: a slim peek card on the Overview page, and the full Forecasting tab in the sidebar.
## Two surfaces
| Surface | Where it lives |
| ------------------ | -------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Overview peek card | The bottom of the [Overview](/en/Systems/Contact-Center-QA/Overview-Page) page. Shows a 2-card KPI pair plus a small chart, plus a **See full forecast** link. |
| Forecasting tab | The dedicated page at `/forecasting`, the last entry in the CCQA sidebar (after TimeLens AI). |
## Page anatomy (Forecasting tab)
| Region | What you see |
| ----------------------- | -------------------------------------------------------------------------------------------------------- |
| Filter bar | Date and hour controls. The hour range matters here, see **The hour filter changes the forecast** below. |
| Forecast controls | Three dropdowns above the KPI strip: **Horizon**, **Training window**, **Confidence band**. |
| Horizon warning | An amber strip that appears only when the chosen horizon is over 60 days. |
| KPI strip | Four cards specific to the forecast. |
| Main chart | A combined chart with historical, forecast, confidence band, and a **Today** marker. |
| Weekday profile chart | Bar chart showing the typical day of the week. |
| Hourly forecast heatmap | A 7-day by 24-hour grid of expected hourly call volumes. |
## Forecast controls
Three dropdowns drive the forecast. The selections are remembered in the URL so a shared link reproduces the same forecast.
| Control | Options | Default |
| ------------------- | ---------------------------------------------------------------------------------------------------------- | ---------------- |
| **Horizon** | **Next 7 days**, **Next 14 days**, **Next 30 days**, **Next 60 days**, **Next 90 days**, **Next 180 days** | **Next 30 days** |
| **Training window** | **Last 28 days**, **Last 56 days**, **Last 84 days**, **All available** | **Last 28 days** |
| **Confidence band** | **20 / 80**, **10 / 90**, **5 / 95** | **20 / 80** |
The **Confidence band** dropdown carries a small help tooltip:
| Confidence band | Meaning |
| --------------- | ------------------------------------------------------------ |
| **20 / 80** | 60% of days fall between the lower and upper bounds. |
| **10 / 90** | 80% of days fall inside the band. Wider, safer for staffing. |
| **5 / 95** | 90% of days fall inside the band. Widest, most conservative. |
## KPI strip
| KPI | Subtitle |
| -------------------------- | ------------------------------------------------------------------------------------- |
| **Next 7 days expected** | Sum of the central forecast for the next 7 days. |
| **Next {n} days expected** | Sum across the chosen horizon. |
| **Peak day in horizon** | The day with the highest central estimate, with the date and the expected call count. |
| **Avg daily (forecast)** | Mean daily forecast across the chosen horizon. |
| **Trained on {n} days** | Footer note, not a card. Tells you how many days of history fed the model. |
The Overview peek card only shows **Next 7 days expected** and **Avg daily (forecast)**.
## Main chart
A combined chart that overlays:
| Series | Visual |
| ------------------- | --------------------------------------------------------------------------------- |
| **Historical** | Solid line, calls per day so far. |
| **Forecast** | Dashed line, the central projection. |
| **Confidence band** | A shaded range from the lower bound to the upper bound of the chosen confidence. |
| **Today** | A vertical reference line where the historical line ends and the forecast starts. |
The **Export chart** button on the card downloads the whole composition as a PNG.
## Weekday profile chart
Mean daily call volume per weekday across the training window.
| Element | What it shows |
| ----------------------- | ------------------------------------------------------------ |
| Title | **Weekday profile**. |
| Subtitle | *Mean daily volume per weekday across the training window.* |
| Bars | One bar per day, ordered Sunday first per product direction. |
| Tooltip | Shows the average call count for that weekday. |
| **Export chart** button | Downloads the bars as a PNG. |
This chart is a sanity check: if the central forecast looks too high or too low, compare against the typical weekday rhythm here.
## Hourly forecast heatmap
A 7 by 24 grid showing expected calls per hour across the next 7 days.
| Element | What it shows |
| ----------------------- | --------------------------------------------------------------------------------------------- |
| Title | **Hourly forecast heatmap**. |
| Subtitle | *Expected calls per hour across the next 7 days.* |
| Grid | Rows are days (next 7 days), columns are hours (00 to 23). Each cell is colored by intensity. |
| Tooltip on cell | Shows the date, the hour, and the expected call count. |
| **Export chart** button | Downloads the heatmap as a PNG. |
The heatmap is rendered to a canvas directly so the PNG export looks identical to the screen.
## Horizon warning
When the chosen horizon exceeds 60 days, an amber strip appears above the main chart with the message *Accuracy degrades beyond 60 days with the seasonal-naive model, upgrade to Holt-Winters or Prophet for longer horizons.*
You can keep the longer horizon, the warning is informational. The forecast continues to project but expect wider real-world variance than the band suggests.
## The hour filter changes the forecast
The page filter bar's hour range scales the forecast. A 09:00 to 17:00 range produces about half the volume of a full-day forecast (roughly 9 hours out of 24). The date range on the filter bar does **not** influence the forecast, the model always projects from the last known data point in your project.
## CSV export
A page-level **Export forecast CSV** button downloads the forecast as a flat table.
| Column | Notes |
| ---------------- | -------------------------------------- |
| `date` | The forecast day. |
| `central` | The central estimate. |
| `lower`, `upper` | Bounds for the chosen confidence band. |
The file uses a UTF-8 BOM and CRLF line endings.
## Common questions
That is by design. The forecast always projects from the last known data point, the date range only affects historical charts (Daily Volume, Hourly Distribution, Heatmap on Overview). Change the **Hour range** on the filter bar to scope the forecast, or use the controls above the KPI strip to change the horizon, training window, or confidence band.
Beyond 180 days, the seasonal-naive model is too coarse to be useful. A future upgrade to Holt-Winters or Prophet will lift the cap. The amber **Horizon warning** strip already advises caution past 60 days.
The main chart shows total calls per day. The heatmap shows calls per hour, so each cell is much smaller than the day's total. Sum a row of the heatmap to compare with the chart.
The peek is hidden when no project is selected, when there are not enough days of history (the seasonal-naive model needs at least one full week per weekday), or when the forecast service has not yet run for this project. The Forecasting tab will surface a clearer state message.
The seasonal-naive model uses the matching weekday's mean. If the training window had a holiday or downtime on, say, every Sunday in scope, the Sunday forecast can collapse near zero. Widen the training window to **All available** to smooth this out.
Per product direction. Most contact centers in the region anchor their week around Sunday rather than Monday.
Yes. The forecast controls are written into the URL (`?fh=`, `?ft=`, `?fb=`) so a shared link reproduces the same view. Browser bookmarks work too.
The main chart and weekday profile use SVG to PNG conversion, which scales to your screen size. If the PNG looks fuzzy, your viewport may be small, resize the browser window before exporting. The heatmap exports directly from a canvas so it always renders at the same crisp resolution.
## Where to go next
Daily volume, hourly distribution, activity heatmap.
Watch call volume drop or spike against a baseline.
Set a target Total Calls value.
Ask "What is the busiest hour of the day?" and get a chart.